[ https://issues.apache.org/jira/browse/HBASE-16417?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15656597#comment-15656597 ]
Eshcar Hillel commented on HBASE-16417: --------------------------------------- Number of regions in all experiments is 50. Figure 3 compares default memstore with mslabs (blue column) and without mslabs (purple column) in a PE run with write only workload uniform distribution. Here are the absolute numbers: || write latencies ||MB/s||50th||75th||95th||99th||99.9th||99.99th||99.999th|| |no compaction|207.2| 57.3| 60.1| 80.7| 9608| 55362| 162462.6| 816825| |no compaction (no mslab)|207.88| 57.0| 59.9| 85.2| 9496| 55521| 167699.7| 933907| Figure 6 compares default memstore with mslabs (blue column) and without mslabs (purple column) in a YCSB run with mixed workload zipfian distribution. Here are the absolute numbers: || read latencies|| op/s|| #gc|| avg|| 50th ||75th|| 95th|| 99th|| |no compaction| 5673| 7693| 3596.0| 2109| 3061.0| 4303| 5535.0| |no compaction (no mslab)| 7900| 6601| 2586| 2137| 3027.0| 4163| 5211.0| I can re-run the benchmarks if you believe this is in contradiction with your previous findings. But the HW is different and might affect the results. > In-Memory MemStore Policy for Flattening and Compactions > -------------------------------------------------------- > > Key: HBASE-16417 > URL: https://issues.apache.org/jira/browse/HBASE-16417 > Project: HBase > Issue Type: Sub-task > Reporter: Anastasia Braginsky > Assignee: Eshcar Hillel > Fix For: 2.0.0 > > Attachments: HBASE-16417-benchmarkresults-20161101.pdf, > HBASE-16417-benchmarkresults-20161110.pdf > > -- This message was sent by Atlassian JIRA (v6.3.4#6332)